I’m testing throughput of a GigaBlox Nano. While other Gigabit botblox switches test out at about 950Mbps, the Nano here is only averaging about 350Mbps, either as a single stream (one port in, one port out) or multiple streams (2 or 3 ports communicating with 1 port). Test setup is the same for both switches.
Has the Nano actually provided Gigabit performance in switching, not just in 1000BASE-T behavior? We need the higher performance for our designs.
Thank you -
For sure, GigaBlox Nano is perfectly capable of 1Gbps throughput, it’s been tested extensively and there’s thousands of units out in customer applications.
The issue is likely something else, my first instinct is that you’re using Jumbo packets (where the size of the packet is more than 1500 kbytes), where GigaBlox Nano can only do standard length packets (<1500Kbytes). It could also be something higher up in the ethernet layers that the switch is not liking, typically
This is a chip level restriction that does not occur on GigaBlox Rugged or UbiSwitch, for example. GigaBlox, however, does share the same chip as GigaBlox Nano, so can you tell me if you also see the same behaviour on GigaBlox?
The only difference between GigaBlox Nano and GigaBlox is that the Nano uses transformerless ethernet (assuming you’re using the PicoConn daughterboard). That could also be the cause.
Tell me if you see the same on GigaBlox, check if you are using jumbo packets, and tell me if you’re using the PicoConn daughterboard, and get back to me.
Hello, Josh,
Thank you for your note. I’m sure the difficulty is not in the ‘Nano, having seen the performance of your products in the past. But just wanted to be certain. Indeed, the latest test used jumbo packets but changing to smaller ones (MTU 1500) has no effect on the throughput.
This is using the PicoConn daughterboard via three “plugable USBC-E2500” NICs on the server laptop and the Surface Pro 4’s docking station 1Gbps port on the client. Both are running Ubuntu 20.4LTE.
I just ran the test on a GigaBlox Rugged – and had exactly the same results. So at this point, I have to assume there’s something odd about the test setup, probably in the Surface Pro 4 end of things. I’ll update the forum post to point that out so there’s no confusion or doubt about the Nano’s performance.
Thank you again, Josh!
Jim Horn l Electrical Engineer
Office: 541.716.4832 l Mobile: 707.327.6253
Email: james@overwatchimaging.com
Web: www.overwatchimaging.com
33 Nichols Parkway, Suite 390, Hood River, OR 97031 USA
Further verification: Windows 11, running iperf3 as administrator for client gets throughput to 949Mbps for the GigaBlox Rugged and the GigaBlox Nano. Excellent! With a 10 volt supply, the Nano takes 115mA or 1.15W. Peak temperature on the board is 71°C in a 25°C environment in still air and no heat sink. With a readily available heat sink, this will do very nicely in the project we picked for it.
Thank you again for another fine product!
Great to hear that. Iperf3 is not deterministic and thus is highly dependent on the load of the CPU is running on.
This is to say, it’s not always a good test; ultimately the CPU is a bottleneck and that’s nothing to do with the network.
For our tests we use two raspberry pis that exclusively run only iperf and we also don’t always see the expected 940mbps.